Tests from reputable youtube channels show significant performance boost in games and loss of 5-10% vs 5800X in almost every other benchmarks.

 

5800X3D performing like 30% better than a 5950X in Ableton? How is that possible? Almost every other test I've seen came to the conclusion that the 5800X3D is worse than even a 5800X for productivity tasks, so how does it beat the 5950X here?

 

Already shared this article in your own thread, that answers that question

 

But in the article you linked, the test in which the 5800X3D performs well is a Kontakt polyphony test, if I understand it correctly. Makes sense it would perform decent in that. But in the DSP test it performs below the 5800X. However, the test that I quoted (the image) was also a DSP test in which they populated each track with various Fabfilter plugins. So unless I'm misunderstanding something, the results contradict each other.

 

Just looked at the test you quoted, yeah, it seems to be DSP related on smaller buffer and something is really off there, like Bitwig and Reaper practically having same performance...

 

Bigger L1-2-3 caches in CPUs have traditionally always been helpful with lower latencies and DSP processing, so no wonder X3D CPUs are better performance wise than X CPUs. However, subjectively, I would still go with a normal X due to its lower power consumption - easier to cool with a 800rpm or less fan quietly and costing 30% less. For me, the difference in performance (like 10% in most cases?) is not justifiably enough to turn a blind eye to these facts. A lot of people won't care about these, that's why I said *subjectively*. It is your choice and in the long run you won't lose whatever you choose. Love these win-win situations.

 

Last edited: May 3, 2023

Well, it doesn't though. Or at least it shouldn't.
According to the article which Melodic Reality linked, the 5800X3D is below even the 5800X in performance when it comes to DSP. And this seems reasonable. But in the DSP benchmark published by a different magazine (which I quoted earlier) it outperforms even the 5950X. It makes sense that it'd maybe perform better in the Kontakt polyphony benchmark, but in the DSP Benchmark it really shouldn't. So I have to assume that there is something very wrong about this benchmark.

 

I grabbed the PDF issue of that magazine and ran it through DeepL. They didn't go into much more detail than this:

 

I just watched an interesting video on YT about CPUs, both AMD and Intel. Rather educative although he is talking about gaming PCs. Gaming is not so different from DAWs and Video editing - they all use CPUs intensively and constantly. Having loads of background tasks running is not desirable as they consume CPU cycles and cause stutters and shit. You can never have a CPU powerful enough that will be able to cope with all the background services and tasks in general that MS or Apple can install and run by default in their OSes. That's why you always have to optimise an OS for audio, video, or gaming.

edit: for the impatient ones, I think last 4 minutes (from 18:25) clear up the situation with CPUs quite nicely. Confusing choices, big price differences, not so much performance differences. I'd still be dancing American Native's war dance around the fire if I had R7 5700G, 5800X or holy spaghetti 7700X! about 250-350€ is a sweet spot to spend on CPUs.

 

Last edited: May 3, 2023

Thanks xorome for doing this!

Still, I can't understand how Reaper ended up with such low scores, did they disabled pre-buffering or whatever is called, even Cubase got better performance and that never was the case, Live and Bitwig are in same ballpark because the way they work, but seen other DAW's that use pre-buffering getting similar results is just strange.

 

Where can I find the PDF version of that issue? I happen to be very fluent in german. Tried googling for the article but couldn't find it.
Anyway, at the end of the day the results from their benchmark still contradict that of other benchmarks, and technically Memory and Cache shouldn't have that big an influence on DSP... For Rompler and Sampler performance, sure, I get it.

Doesn't help that their article doesn't really explain why the cache would help in DSP, they just seem to be assuming that that's the thing that's making the difference. Idk, very strange. Did they maybe run the DSP test on tracks running Kontakt instruments or something? They're mentioning the processing plugins they put on the tracks, but not the sound source.

 

Last edited: May 3, 2023

Is 3700 not enough???? Do you have a 100% loaded when playing all tracks? What music do you write?

 

Tbh i think that if you can afford it, getting the 5900x would be a financial decision you'd get the most longevity out of as not only are the boost clocks higher, (better single core performance as such) the higher core counts would give you much better parallel dsp power, with only truly massive projects bottlenecking your performance. I think it does everything very very well. The 5800x3d would be worth it if your main tools happen to be large sample libraries like Omnisphere or Kontakt tho, as this is where the increased cache really gets a chance to stretch it's legs. The 5900x is the best of all worlds tho.

 

It is not about music! It is about big balls for less money

 

https://www.heise.de/select/ct/2022/12/seite-22

 

I actually own the 5800X.
Its DSP-performance is more than sufficient even using several DSP-56300 emulations simultaniously is not an issue.
Including Ostiris...
The 5800X and the 5800X3D have a monolithic 8 Core-CCD/CCX, whereas the 12-Core 5900X probably introduces additional latency as 2 CCD/CCX with 6 cores each are used.
The communication-overhead using the Infinity fabric (scheduling, cache access) for sure has a negative impact regarding DPC-latency.

 

Last edited: Apr 26, 2024

Super interesting. Thank you so much for this hint with the splitted CPUs!!

What do you think about this guys for serious Ostirus Emulation in FL and Ableton + many FX:

AMD Ryzen 7 7700X 8x 4.50GHz (32 MB L3 Cache)
OR
AMD Ryzen 7 7800X3D 8x 4.20GHz (96MB L3 Cache!! )

I heared that the massive L3 Cache can help with the latency.

I also want to play some AAA Games and want to install a Hackintosh together with my RX580

Thank you

 

Yeah the n900X variant will finally be worth it.

 

They didn't double the number of cores though! Maybe wait for Zen 6?

 

No, in fact I don't know what I thought I read when I replied, but I do know that it will be 16 instead of 12, which is why I think i will be worth it.